SEGMENTATION OF COLOR BUILDING IMAGES BASED ON WATERSHED AND REGION MERGING

A kind of color building images segmentation method was proposed here. This approach applied area control pretreatment and the watershed algorithm to produce the original regions. The axial symmetry combined with other properties was estimated as region similarity features. The final segmentation was derived by using merging process. And the termination criterion was determined from the distribution (histogram) of the merging costs as an adaptive threshold. The results show that the approach proposed here can produce effective region of building and reduce the noise influence to segmentation. So it's very significant for building image segmentation in 3D reconstruction.